Investigations in Neuromodulation

Journal of Neurotherapy: Investigations in Neuromodulation, Neurofeedback and Applied Neuroscience Induced EEG Gamma Oscillation Alignment Improves Differentiation Between Autism and ADHD Group Responses in a Facial Categorization Task Eric Gross a , Ayman S. El-Baz a , Guela E. Sokhadze a , Lonnie Sears b , Manuel F. Casanova a c a c & Estate M. Sokhadze a Department of Bioengineering , University of Louisville , Louisville , Kentucky , USA b Department of Pediatrics , University of Louisville School of Medicine , Louisville , Kentucky , USA c Department of Psychiatry and Behavioral Sciences , University of Louisville School of Medicine , Louisville , Kentucky , USA Published online: 29 May 2012.

To cite this article: Eric Gross , Ayman S. El-Baz , Guela E. Sokhadze , Lonnie Sears , Manuel F. Casanova & Estate M. Sokhadze (2012) Induced EEG Gamma Oscillation Alignment Improves Differentiation Between Autism and ADHD Group Responses in a Facial Categorization Task, Journal of Neurotherapy: Investigations in Neuromodulation, Neurofeedback and Applied Neuroscience, 16:2, 78-91, DOI: 10.1080/10874208.2012.677631

To link to this article: http://dx.doi.org/10.1080/10874208.2012.677631

PLEASE SCROLL DOWN FOR ARTICLE

© International Society for Neurofeedback and Research (ISNR), all rights reserved. This article (the “Article”) may be accessed online from ISNR at no charge. The Article may be viewed online, stored in electronic or physical form, or archived for research, teaching, and private study purposes. The Article may be archived in public libraries or university libraries at the direction of said public library or university library. Any other reproduction of the Article for redistribution, sale, resale, loan, sublicensing, systematic supply, or other distribution, including both physical and electronic reproduction for such purposes, is expressly forbidden. Preparing or reproducing derivative works of this article is expressly forbidden. ISNR makes no representation or warranty as to the accuracy or completeness of any content in the Article. From 1995 to 2013 the Journal of Neurotherapy was the official publication of ISNR (www. Isnr.org); on April 27, 2016 ISNR acquired the journal from Taylor & Francis Group, LLC. In 2014, ISNR established its official open-access journal NeuroRegulation (ISSN: 2373-0587; www.neuroregulation.org).

THIS OPEN-ACCESS CONTENT MADE POSSIBLE BY THESE GENEROUS SPONSORS

INDUCED EEG GAMMA OSCILLATION ALIGNMENT IMPROVES DIFFERENTIATION BETWEEN AUTISM AND ADHD GROUP RESPONSES IN A FACIAL CATEGORIZATION TASK

Eric Gross1, Ayman S. El-Baz1, Guela E. Sokhadze1, Lonnie Sears2, Manuel F. Casanova1,3, Estate M. Sokhadze1,3 1Department of Bioengineering, University of Louisville, Louisville, Kentucky, USA 2Department of Pediatrics, University of Louisville School of Medicine, Louisville, Kentucky, USA 3Department of Psychiatry and Behavioral Sciences, University of Louisville School of Medicine, Louisville, Kentucky, USA

Children diagnosed with an autism spectrum disorder (ASD) often lack the ability to recognize and properly respond to emotional stimuli. Emotional deficits also characterize children with attention deficit/hyperactivity disorder (ADHD), in addition to exhibiting limited attention span. These abnormalities may effect a difference in the induced EEG gamma wave burst (35–45 Hz) peaked approximately 300–400 ms following an emotional stimulus. Because induced gamma oscillations are not fixed at a definite point in time poststimulus, analysis of averaged EEG data with traditional methods may result in an attenuated gamma burst power. We used a data alignment technique to improve the averaged data, making it a better representation of the individual induced EEG gamma oscillations. A study was designed to test the response of a subject to emotional stimuli, presented in the form of emotional facial expression images. In a four-part experiment, the subjects were instructed to identify gender in the first two blocks of the test, followed by differentiating between basic emotions in the final two blocks (i.e., anger vs. disgust). EEG data were collected from ASD (n ¼ 10), ADHD (n ¼ 9), and control (n ¼ 11) subjects via a 128-channel EGI system, and processed through a continuous wavelet transform and bandpass filter to isolate the gamma frequencies. A custom MATLAB code was used to align the data from individual trials between 200 and 600 ms poststimulus, EEG site, and condition by maximiz- ing the Pearson product–moment correlation coefficient between trials. The gamma power

for the 400-ms window of maximum induced gamma burst was then calculated and compared between subject groups. Condition (anger/disgust recognition, gender recognition) Alignment Group (ADHD, ASD, Controls) interaction was signiﬁcant at most of parietal topographies (e.g., P3-P4, P7-P8). These interactions were better manifested in the aligned data set. Our results show that alignment of the induced gamma oscillations improves sensitivity of this measure in differentiation of EEG responses to emotional facial stimuli in ADHD and ASD.

Received 27 December 2011; accepted 16 February 2012. Funding for this work was provided by the National Institutes for Health grant R01 MH86784 to Manuel Casanova. Address correspondence to Estate (Tato) M. Sokhadze, PhD, Cognitive Neuroscience Laboratory, Department of Psychiatry & Behavioral Sciences, University of Louisville School of Medicine, 401 East Chestnut Street, Suite 600, Louisville, KY 40202, USA. E-mail: [email protected]

78 INDUCED EEG GAMMA ALIGNMENT IN FACIAL TASK 79

INTRODUCTION gamma frequencies, particularly those cen- tered about 40 Hz, have been tied to visual, Autism spectrum disorder (ASD) and attention attentional, cognitive, and memory processes deficit=hyperactivity disorder (ADHD) are both (Bas¸ar, Schu¨rmann, Bas¸ar-Eroglu, & Demiralp, early onset neurodevelopmental disorders. 2001). Following a stimulus, two gamma oscil- Children with autism are typically character- lations are typically noted: an early evoked ized by social and emotional deficits stemming oscillation and a late induced oscillation from an inability to properly perceive and (Bas¸ar-Eroglu, Stru¨ber, Schu¨rmann, Stadler, & respond to these forms of stimuli. In children Bas¸ar, 1996; Bas¸ar, Schu¨rmann, Bas¸ar-Eroglu, with ADHD, attentional, impulse, and motor & Demiralp, 2001). The evoked gamma oscil- controls are notably decreased. Although most lations typically occur within the first 200 ms studies have typically separated these con- after the onset of a stimulus and are locked ditions as unrelated phenomena, more recent reviews have justified the comparison of these in time from trial to trial. Because little variation disorders in a combined experiment, which is seen in the latency of the evoked gamma may potentially reveal mechanistic differences with changing stimulus type, it is believed that between similar behaviors in ASD and ADHD it may be a result of sensory processes. Con- (Rommelse, Geurts, Franke, Buitelaar, & versely, induced gamma oscillations occur Hartman, 2011). later, after 240-ms poststimulus and vary in The theory of mind (ToM) represents the latency from trial to trial (Tallon-Baudry & attempt of assuming another’s perspective by Bertrand, 1999). These variations may suggest characterizing their mental state, or comparing that the induced gamma oscillations are related it to one’s own (Baron-Cohen, 2000; to higher cognitive processes (Tallon-Baudry, Baron-Cohen & Belmonte, 2005). The ToM 2003). Deviations from typical gamma band construct is frequently applied in the study of activity have been reported in several studies ASD (Ahmed & Miller, 2011; Lerner, Hutchins, on neurological and psychiatric disorders, inc- & Prelock, 2011; Sabbagh, 2004) and may luding epilepsy, Alzheimer’s disease, ADHD, explain why autistic children struggle with and ASD (Herrmann & Demiralp, 2005). understanding facial expressions, body lan- Because evoked gamma waveforms are guage, figurative speech, and other social cues synchronized in time poststimulus, averaging that convey emotional information. Applica- analogous trials typically reveals the evoked tions of this theory have been used to assess response in the averaged waveform. However, both the nature and level of emotional defi- induced gamma waveforms vary in time, and ciencies in adults with ASD (Baron-Cohen, thus are not typically represented in the aver-

Wheelwright, & Jolliffe, 1997). Applying ToM aged signal. This makes the analysis of the to other conditions, such as ADHD (Buhler, induced gamma waveform more complex than Bachmann, Goyert, Heinzel-Gutenbrunner, & evoked gamma waveforms. Thus, studies look- Kamp-Becker, 2011), may provide a perspec- ing at averaged oscillatory gamma waveforms tive that allows for a better understanding of have either focused on evoked gamma (Lenz the disorder. et al., 2008) or used time-frequency analysis A quantitative analysis of the electroence- to ﬁnd and characterize induced gamma phalographic (EEG) data during a ToM task in activity (Mu¨ller, Gruber, & Keil, 2000). subjects with neurodevelopmental disorders Data alignment is a procedure that corre- (e.g., ASD, ADHD) provides a top-down lates analogous features between two signals approach to better correlate physiological and or images, and standardizes them so they behavioral responses. EEG oscillations are may be more compared or analyzed to one separated into several frequency bands, ran- another (Figure 1). Data alignment has been ging from the slower delta waves (0–4 Hz) to used in other studies to align visual evoked the faster gamma waves (30–80 Hz). The potentials with varying latencies via the 80 E. GROSS ET AL.

FIGURE 1. An example of the effect of data alignment with simplified impulse signals. Note. The first trial within a set is used as an alignment setpoint for subsequent trials. Averaging the signals with data alignment produces a representative signal that resembles the con- stituent trials, whereas the unaligned averaged signal is significantly attenuated. (Color figure available online.) discrete Fourier transform (Sahin & Yilmazer, Diagnosis was made according to the Diagnostic 2010). With induced gamma waveforms in and Statistical Manual of Mental Disorders EEG, a similar technique may be used to align (4th ed., text rev. [DSM–IV–TR]; American Psy- the induced gamma ‘‘burst’’ that occurs after chiatric Association [APA], 2000) and further the evoked gamma activity. ascertained with the Autism Diagnostic Inter- This study proposes a novel method of view–Revised (LeCouteur, Lord, & Rutter, visualizing the induced gamma activity of an 2003). They also had a medical evaluation by averaged EEG response through a method of a developmental pediatrician. All subjects had data alignment, which may allow for a more normal hearing based on past hearing screens. accurate representation of the averaged Participants either had normal vision or wore induced gamma activity of a subject. This pro- corrective lenses. Participants with a history of cess may contribute to the analysis of emotion- seizure disorder, significant hearing or visual al and attentional differences in ADHD and impairment, a brain abnormality conclusive

ASD. Our hypothesis was that differences from imaging studies or an identiﬁed genetic between ADHD, ASD, and control subjects disorder were excluded. All participants would manifest themselves in the power values with autism were high-functioning persons with of the induced gamma relatively better when Full Scale IQ greater than 80 assessed using the processed with a data alignment technique. Wechsler Intelligence Scale for Children, Fourth Stimuli in the form of facial images, both Edition (Wechsler, 2003) or the Wechsler expressive and nonexpressive, would be used Abbreviated Scale of Intelligence (Wechsler, as a means of producing gamma oscillations. 1999). The Structured Clinical Interview for METHODS DSM–IV (SCID-I=P; First, Spitzer, Gibbon, & Williams, 2001a) was used for diagnoses of Subjects ADHD. Nine subjects aged 13 to 21 who cur- Participants with ASD (age range ¼ 9–20 years) rently meet DSM–IV–TR criteria for ADHD or were recruited through the University of ADD (APA, 2000) were included. Subjects Louisville Weisskopf Child Evaluation Center. were evaluated at the Weisskopf Child INDUCED EEG GAMMA ALIGNMENT IN FACIAL TASK 81

Evaluation Center. To confirm the diagnosis of and the mean age of the ADHD group was ADHD parents and teachers completed the 14.2 3.9 years (N ¼ 9, range ¼ 10–19 years, Child Behavior Checklist or Teacher Report seven male, two female). The mean age of Form (Achenbach & Rescorla, 2001). Parents the control group (N ¼ 11) was 14.8 4.5 were also interviewed using DSM–IV criteria years (9–21 years, eight male, three female). for ADHD to confirm the diagnosis. Only sub- The age difference between groups was not jects with clinical features meeting criteria for significant. The Mean Full Scale IQ scores were ADHD in both the home and school setting 94.2 18.1 for patients with ASD and and who meet DSM criteria were included. 98.6 9.2 for children with ADHD. Six sub- All ADHD participants had a medical history jects from the ADHD group and six subjects and a psychiatric evaluation (for children, both from the ASD group were on medication. parents and children provided information for The children with ADHD were taking stimu- the assessment). lants (Methylphenidate or Dextroampheta- Controls were recruited through advertise- mine). Two children with ASD were also ments in the local media. All control parti- taking stimulants (Concerta, Adderall), and four cipants were free of neurological or significant were taking antidepressants (Fluoxetine, Sertra- medical disorders, had normal hearing and line) and mood stabilizers (Divalproex, Aripra- vision, and were free of psychiatric, learning, zole). Two children in the ASD group had or developmental disorders based on self- comorbid mild mood disorders, and two had and parent reports. Subjects were screened co-occurring anxiety disorders. One subject for history of psychiatric or neurological diag- from the ADHD group had comorbid mild nosis using the SCID Non-Patient Edition mood disorder, and one had anxiety disorder. (First, Spitzer, Gibbon, & Williams, 2001b). Participants within the control, ADHD, and Gender/Emotion Recognition Task autism groups were attempted to be matched Stimulus presentation for the gender=emotion by age, Full Scale IQ, and socioeconomic status recognition task was controlled via the E-prime of their family. Socioeconomic status of ASD, software (Schneider, Eschman, & Zuccolotto, ADHD, and control groups was compared 2002). Facial images were displayed on a based on parent education and annual house- 15-in. flat-panel display. Subjects were seated hold income. Participants in the three groups during the study, and a chinrest was provided had similar parent education levels. to keep the center of the display approximately Participants and their parents (or legal 50 cm from the subject’s eyes. Subject guardians) were provided with full information responses were collected via a keypad (Serial about the study including the purpose, Box, Psychology Software Tools, Pittsburgh, requirements, responsibilities, reimbursement, PA). Instructions varied between the four risks, benefits, alternatives, and role of the blocks of the study and were presented on local Institutional Review Board. The consent the screen to the subject prior to beginning a and assent forms approved by the Institutional new segment of the test. All four segments Review Board were reviewed and explained required the subject to select one of two to all subjects who expressed interest to par- choices by pressing the corresponding button ticipate. All questions were answered before on the keypad explained in the instructions. consent signature was requested. If the indi- EEG signals recorded using a 128-channel vidual agreed to participate, she or he signed Electrical Geodesics Inc. system (EGI; Eugene, and dated the consent form and received a OR) were sampled at 500 Hz and passed copy countersigned by the investigator who through an analog bandpass filter (0.1– obtained consent. 200 Hz) and referenced to the vertex at Cz. The mean age of 10 participants enrolled The Geodesic Sensor Net is a lightweight, elas- in the ASD group was 14.1 (SD) 2.7 years tic structure housing the silver=silver-chloride (range ¼ 10–18 years, eight male, two female), electrodes within a synthetic sponge on a 82 E. GROSS ET AL. pedestal. Sponges were soaked in potassium differentiating either the gender or the per- chloride prior to testing to promote conducti- ceived emotional facial expression in the vity. Sensor impedance was maintained below image. The subjects indicated that difference recommended manufacturer specification of by pressing the corresponding button on the 40 kX. Collected signals were segmented off- keypad. Each category contained 60 images line into one second trials ranging from for the subject to differentiate. Images 200-ms prestimulus to 800-ms poststimulus. remained on the screen for 300 ms, whereas Facial images were organized into four EEG recording occurred for a full 1-s period. categories: (a) gender recognition with neutral Pauses between stimuli ranged from 1,100 to expressions, (b) gender recognition with 1,300 ms to avoid anticipatory effects. The emotional expressions, (c) anger versus disgust complete four category experiment took recognition, and (d) fear versus sadness recog- approximately 20 min to complete, including nition (Figure 2). Each category contained 24 short breaks that were provided between unique images, with equal representation of image categories, presentation of the instruc- male and female subjects. Similarly, in emotions, and brief practice sessions prior to each tion recognition tasks, each emotion was category. equally represented. Seventy-two total images Collected signals were stored in EGI Net were used for all four categories, with some Station, tagged according to test category, and reuse between categories. All images were segmented into 1-s trials. The data were then randomly selected from standard databases organized into four experimental conditions of facial pictures (Pictures of Facial Affect, by based on the task the subject was asked to per- Ekman 1976–2004, Berkeley, CA; JACFEE= form: (a) Gender Recognition-All, (b) Emotion JACNeuF, Matsumoto & Ekman, 1988–2004, Recognition-All, (c) Anger=Disgust Recognition, Berkeley, CA). and (d) Fear=Sad Recognition (Figure 3). Eleven The experiment was divided into four sec- posterior (parietal, parieto-occipital, and tions, corresponding to the four categories of occipital EEG sites including P3, P4, P7, P8, facial images described earlier. In each section, P9, P10, PO3, PO4, POz, O1, and O2 acc- the subject was asked to identify the displayed ording to 10-10 International System) were faces as belonging to one of two groups, selected for induced gamma power analysis. Posterior channels were reported to be

FIGURE 2. A representation of the four block experimental study and examples of the facial images used during the test procedure. Participants were asked to distinguish a face as belonging to one of two groups: male or female, angry or disgusted, or fear- FIGURE 3. The four experimental categories used for data analy- ful or sad. Each test block consisted of 60 trials, with 24 unique sis. For each subject, 60 trials were selected for analysis in the images per trial. Facial stimuli were presented at the screen for gender and overall emotion recognition categories, whereas 30 300 ms with the intertribal interval varying in 1,100–1,300 ms. were selected for the individual emotion recognition tasks (i.e., (Color figure available online.) anger vs. disgust). (Color figure available online.) INDUCED EEG GAMMA ALIGNMENT IN FACIAL TASK 83 preferable EEG sites during visual stimulation Subsequent trials in the group were then for induced gamma analysis (Mu¨ller et al., compared to the setpoint. For each trial, a 2000). Approximately 30 trials were used for 400-ms window starting at 100-ms poststimu- analysis in the Anger=Disgust and Fear=Sad rec- lus was initially selected (i.e., 100 to 500 ms ognition for each subject, and 60 trials were poststimulus). The two-dimensional Pearson used in the Gender=Emotion Recognition. Product–Moment correlation coefficient These data were exported into MATLAB (Equation 2) was then calculated between this (Math Works Inc., Natick, MA) for further signal window and the setpoint. processing. P P ÀÁ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffim n Amn AA ðÞBmn BB Filtering Technique r ¼ P P P P 2 2 ð ðAmn AAÞ Þð ðBmn BBÞ Þ EEG data collected from the gender=emotion m n m n recognition task were first processed via wavelet analysis. This technique allows for visualiza- The window was shifted by 2 ms forward in tion of the collected signals in both the time time (i.e., 102 to 502 ms poststimulus) and the and frequency domains, providing information coefficient calculation was repeated. This pro- about the amplitude of gamma waveforms at cess was performed iteratively 101 times, shift- varying frequencies within the selected time ing the window incrementally to cover a total interval. A one-dimensional continuous wave- time span of 100–700 ms poststimulus. The let transform (Equation 1) was performed using 400-ms window with the largest correlation the MATLAB Wavelet Toolbox. coefficient was then selected as the ‘‘aligned’’ form of the signal and was exported into a Z database of aligned data (Figure 4). This pro- w 1 t s CWT ðÞ¼s; S pffiffiffi xtðÞw dt cess was repeated for all signals within a group, x S S and for all groups in the original data set. An unaligned database was also created by simply The Morlet window was selected as the segmenting the original trials from 200 to 600 mother wavelet in this analysis. One hundred ms poststimulus. twenty-eight wavelet coefficients were found Trials within each group were averaged from each signal. Following wavelet analysis, together in MATLAB to produce a 400-ms sig- a custom Harris bandpass filter was applied nal for both the aligned and unaligned data to the signals to isolate frequencies of interest. sets. Gamma power was calculated by sum- This filter allowed for the passage of the gamma ming the squares of the amplitude at each frequencies between 35–45 Hz with a 2-Hz point in the averaged signals. attenuation band. A similar Wavelet=Harris filtering technique was used in previous gamma Statistical Analysis analysis studies on neurofeedback and cue Data analysis was performed in SPSS (v. 18) reactivity (Horrell et al., 2010). and MINITAB (v. 16) statistical software packages. Gamma power values calculated in Data Alignment and Averaging the previous step were loaded into the pro- Filtered data were processed further in gram following the removal of outliers. A MATLAB to create an aligned data set. Seg- repeated measures analysis of variance mented trials were organized into groups by (ANOVA) was performed with a combination subject, experimental condition, and EEG of the following factors: Subject Group (ADHD, channel. Within each group, the first trial was ASD, or control), Experimental Condition selected as a setpoint for alignment. A (Anger=Disgust, Fear=Sad, etc.), Channel (P3, 400-ms window from 200 to 600 ms poststi- P4, etc.), Hemisphere (right or left), and mulus was then segmented from the trial to Alignment (aligned and unaligned). Models capture the induced gamma activity. were constructed to test for significant 84 E. GROSS ET AL.

FIGURE 4. The step-by-step procedure of the alignment technique. First, the setpoint is chosen by segmenting the first signal in a set from 200–600 ms poststimulus. Subsequent trials are then incrementally segmented in 400-ms pieces starting at 100-ms poststimulus, with a 2-ms shift each increment. The correlation coefficient is calculated for each increment, and the level of shift that produces the highest coefficient value is selected as the ‘‘aligned’’ 400-ms segment for analysis. This process is repeated for each trial within a set until each trial is aligned to the setpoint. (Color figure available online.)

interactions between-subject group, experi- main effect for alignment was observed mental condition, hemisphere, and alignment individually in all channels and hemispheric for channel pairs (i.e., P3 and P4, P7 and P8, channel pairs. Graphs prepared in MATLAB etc.). Experimental conditions varied in our allowed for visualization of the alignment effect ANOVA models. Simple models compared in the averaged signals for subject, experi- the gender and emotion recognition tasks gen- mental condition, and channel pairings erally (i.e., Gender All vs. Emotion All) while (Figure 5). more speciﬁc models looked at the individual Signiﬁcant Group Condition Alignment emotion recognition tasks separately and com- three-way interactions were observed gener- pared them to the gender recognition task (i.e., ally across the parietal and occipital channels

Anger=Disgust vs. Gender All). Greenhouse- using a model that compared the separate Geisser corrected p-values were used for emotion recognition tasks and the gender rec- determination of statistical significance when ognition task (F ¼ 2.68, p ¼ .03). In models appropriate. that compared the Anger=Disgust recognition task to the gender recognition task, significant interactions could be seen in the P3-P4 chan- RESULTS nels (F ¼ 3.43, p ¼ .048) and P7-P8 channels A significant main effect of alignment (F ¼ 4.30, p ¼ .025). As shown in Figures 6 (F ¼ 995.89, p < 0.0001) was observed across and 7, significant effects of Condition Group all parietal, parieto-occipital, and occipital Group interaction became more apparent in channels collected (P3, P4, P7, P8, P9, P10, the aligned data sets. The interaction effect POz, PO3, PO4, O1, O2, see Table 1). This in the aligned data set can be described effect was observed in all ANOVA models as a higher power of induced gamma in regardless of the experimental conditions sel- ADHD group as compared to ASD group dur- ected for comparison. Similarly, the significant ing anger versus disgust differentiation task, INDUCED EEG GAMMA ALIGNMENT IN FACIAL TASK 85

TABLE 1. Analysis of Variance Table for Induced Gamma Power with Condition, Group, and Alignment Main Effects and Interactions in Parietal=Occipital Channels

Factor Type Levels Values

Condition Fixed 3 Anger-Disgust, Fear-Sad, Gender-All Alignment Fixed 2 Aligned, Unaligned Group Fixed 3 ADHD, Autism, Control Analysis of Variance for Power-P, using Adjusted SS for Tests

Source DF Seq SS Adj SS Adj MS F P

Condition 2 11.41 11.52 5.76 3.45 0.032 Alignment 1 1670.95 1663.13 1663.13 995.89 0.000 Group 2 80.12 80.03 40.01 23.96 0.000 Condition-Alignmebt 2 2.64 2.52 1.26 0.76 0.470 Condition-Group 4 2.39 2.39 0.60 0.36 0.839 Alignment-Group 2 0.98 1.04 0.52 0.31 0.733 Condition-Alignment-Group 4 17.91 17.91 4.48 2.68 0.030 Error 1824 3046.07 3046.07 1.67 Total 1841 4832.46

FIGURE 5. An example of aligned and unaligned EEG signals at parietal channels (P7, P8) for a single subject. (Color ﬁgure available online.)

FIGURE 6. Signiﬁcant interaction plots for parietal channels P3 and P4 depicting differences in condition, group, and alignment pairings. (Color ﬁgure available online.) without any differences during gender recog- DISCUSSION nition task at the parietal sites (P3, P4, P7, Effect of Alignment P8). Descriptive statistics for P3-P4 and P7-P8 three-way interaction groups are pro- The extremely large main effect of alignment vided in Table 2. may be explained by the nature of the 86 E. GROSS ET AL.

FIGURE 7. Signiﬁcant interaction plots for parietal channels P7 and P8 depicting differences in condition, group, and alignment pairings. (Color ﬁgure available online.)

MATLAB program utilized in this study. Our channels suggest that the alignment procedure program aligned trials by shifting them within produces data that better resolves the differ- a fixed window of time to maximize the ences between group-condition pairings. amount of overlap that occurs. This reduced Whereas significant group-condition effects the attenuation of the averaged signal. If the would have gone unnoticed in the parietal maximum overlap hypothetically occurred channels with traditional techniques, align- from trial to trial before performing any shift ment provided a means of visualizing these sig- on the time axis, the ‘‘aligned’’ data set would nificant differences between ADHD, ASD and be identical to the ‘‘unaligned’’ data set. Thus, control subjects, as shown in Figure 6. the power of the aligned averaged waveform The outlined alignment procedure may be should always be equal to or greater than the modified for future studies. A wide 400-ms power of the unaligned averaged waveform, window was selected to ensure that the as the program will not produce an aligned sig- induced gamma region of the signals was cap- nal that is more attenuated than the unaligned tured, though this window could be changed signal. This effect was confirmed visually by to any value less than the total length of the sig- examining the graphs of aligned and unaligned nal. Similarly, the selection of the setpoint waveforms produced in MATLAB, as shown in window from 200 to 600 ms poststimulus Figure 5. could be shifted if the induced gamma is antici- Significant Group Condition Alignment pated to occur at a different point in time. The three-way interactions seen in the parietal incremental comparisons between the setpoint

TABLE 2. Descriptive Statistics for the Parietal EEG Channels

Aligned Unaligned

Patient group Experimental condition Mean Standard error Mean Standard error

P3-P4 ADHD Anger=Disgust 2.049 0.286 0.450 0.126 ADHD Gender All 1.345 0.203 0.315 0.084 Autism Anger=Disgust 1.292 0.283 0.320 0.126 Autism Gender All 1.326 0.203 0.143 0.084 Control Anger=Disgust 0.922 0.271 0.129 0.120 Control Gender All 0.943 0.192 0.072 0.080 P7-P8 ADHD Anger=Disgust 3.355 0.500 0.987 0.215 ADHD Gender All 2.798 0.491 0.620 0.160 Autism Anger=Disgust 2.031 0.530 0.333 0.228 Autism Gender All 2.769 0.520 0.172 0.169 Control Anger=Disgust 2.697 0.474 0.418 0.204 Control Gender All 2.524 0.465 0.181 0.151

Note. All pairings between aligned and unaligned samples were statistically significant (p < .02) with a Welch’s t test. ADHD ¼ attention deficit=hyperactivity disorder. INDUCED EEG GAMMA ALIGNMENT IN FACIAL TASK 87 and subsequent trials in a group were made According to Bachevalier and Loveland every 2 ms based on the system sampling fre- (2006), early dysfunctions in a complex of quency of 500 Hz, though this value could be neural structures involved in social cognition, increased to improve the speed of the program which includes the ventromedial parts of the at the cost of lower resolution. The time range prefrontal cortex, the amygdala within tem- examined in the incremental comparisons was poral lobe, and their interconnections with set from 100 to 700 ms poststimulus, but this the other limbic structures and brainstem, range may be changed as needed. may result in a severe impairments in facial A potential source of error this alignment expression recognition, and in understanding technique introduces is the selection of a set- of other socially meaningful gestures leading point. In this study, the first trial in each to profound deficiency in awareness of the Subject-Condition-Channel group was seg- social significance of emotional stimuli and mented from 200–600 ms and used to align situations. Considering that the children with the subsequent trials in the group. If this trial neurodevelopmental disorders in our study had artifacts or grossly abnormal induced were high-functioning individuals, they might gamma activity, it is possible that the system have relatively spared ability to recognize rela- may align the subsequent trials improperly. tively simple emotional expressions (e.g., sad= This may be alleviated by examining trials prior fear pair), though recognition of more complex to analysis, as was done in this study. Future negative emotions (e.g., anger=disgust pair) efforts may include incorporating an algorithm was still impaired, especially in children with that examines the setpoint prior to alignment, ASD. It was reported repeatedly that for indivi- and accepts or rejects it based on user- duals with autism it is harder to decode contributed criteria (i.e., amplitude threshold, emotional expressions of faces (Baron-Cohen, minimum power, etc.). 2000; Sabbagh, 2004; Sabbagh, Moulson, & Harkness, 2004; Schultz, 2005) and more diffi- Comparisons of Induced Gamma Power cult to detect a difference between two in Subject Groups emotional facial expressions (Ashwin, Wright, Prior to analysis, it was hypothesized that the & Baron-Cohen, 2006). emotion recognition tasks would be more chal- A meta-analysis of the ToM experiments lenging for ADHD and ASD subjects than the (U. Frith, 2001) highlighted involvement of a gender recognition tasks and would be more periamygdalar area of temporal lobe, a para- likely to affect changes in the induced gamma cingulate area of the medial frontal cortex, waveforms between the subject groups. The and areas between temporal and parietal cor- significant interactions in Figure 6 reveal some tices in mentalizing. A review by Schultz trends that support this hypothesis. Within the (2005) made a specific focus on face percep- aligned data sets, the power of the gender rec- tion deficits in autism describing extensive ognition task remained relatively constant neuroimaging literature on abnormalities in between ASD, ADHD, and control subjects. the fusiform face area, concluding that indivi- Much greater variation is seen in the anger= duals with ASD are selectively impaired in their disgust recognition task. ADHD subjects typi- ability to recognize faces and concomitantly cally exhibited a higher induced gamma power differentiate emotional facial expressions. during this task compared to the gender recog- Reduced induced gamma activation of parietal nition task. Conversely, ASD subjects had a and parieto-occipital cortices in our study is in lower induced gamma in the anger=disgust accord with repeatedly demonstrated reduced recognition task versus the gender recognition activity in these cortical regions in individuals task. Control subjects had relatively small dif- with ASD (Castelli, Frith, Happe, & Frith, 2002; ferences between the induced power of the Di Martino & Castellanos, 2003; C. D. Frith & two tasks compared to ADHD (P3-P4 and Frith, 1999; Greimel et al., 2010). Brock and P7-P8) and ASD (P7-P8) subjects. colleagues (Brock, Brown, Boucher, & Rippon, 88 E. GROSS ET AL.

2002; Rippon, Brock, Brown, & Boucher, showed that evoked (Baruth et al., 2010) and 2007) proposed that underconnectivity induced gamma power (Sokhadze et al., between separate functional brain regions, 2009) in autism improved following neuromo- including those involved in social cognition, dulation treatment using repetitive transcranial might be reflected in a lack of coactivation of magnetic stimulation, development of more EEG activity in gamma band. In neurotypical reliable functional outcome such as aligned subjects, induced gamma activity is modulated induced gamma measures is definitely an by a various integrative processes (Belmonte important and feasible goal. et al., 2004), such as feature binding (Gray & Singer, 1989; Llinas & Steriade, 2006; Singer, CONCLUSION 1999; Tallon-Baudry, 2003), attention (Mu¨ller et al., 2000), face processing (Rodriguez et al., A data alignment technique was used to 1999), emotion (Keil et al., 2001), and memory improve representation of the individual rehearsal (Tallon-Baudry & Bertrand, 1999). induced EEG gamma oscillations in groups of Decreased and delayed induced gamma oscil- children with autism, ADHD, and typical con- lations in visual processing areas in our facial trols. A study was designed to analyze group emotion recognition task may suggest disrup- differences in induced gamma responses to tion of neural signaling and is supportive of emotional stimuli in the form of emotional the hypothesis of abnormal regional gamma facial expression images. The gamma power activation patterns in autism. for the window of maximum induced gamma In typically developing young children, the burst was calculated and compared between face holds specific significance and provides three groups. Condition (anger=disgust recog- vital nonverbal information important for nition, gender recognition) Alignment communication (Dawson, Webb, Carver, Group (ADHD, ASD, controls) interaction was Panagiotides, & McPartland, 2004). ERP stu- significant at most of parietal EEG sites. These dies in typical and idiopathic developmentally interactions were better manifested in the delayed infants showed that young children aligned data set, especially when autism and with autism have an impaired ability to process ADHD group were compared. Our results faces (Dawson et al., 2002; de Haan & Nelson, show that alignment of the induced gamma 1999; Deruelle, Rondan, Gepner, & Tardif, oscillations improves sensitivity of this measure 2004). The ability to understand and use facial in differentiation of EEG responses to emotion- emotional information is one of the core defi- al facial stimuli in ADHD and ASD. cits in autism (Baron-Cohen, 2000), and separ- ation of deficits in general facial processing REFERENCES (e.g., recognition of the gender of shown face) from deficits in recognition and understanding Achenbach, T. M., & Rescorla, L. A. (2001). specific emotional facial expressions (differen- Manual for the ASEBA school-age forms & pro- tiation of angry vs. disgust facial expressions) files: child behavior checklist for ages 6–18, is a relatively difficult task that requires very teacher’s report form, youth self-report, An sensitive quantitative EEG measures and com- integrated system of multi-informant assess- parison of outcomes of two subtasks (gender ment. Burlington: University of Vermont, vs. specific emotion recognition). Comparative Research Center for Children, Youth, and analysis of induced gamma responses during Families. facial recognition task with two contrast groups Ahmed, F. S., & Miller, L. S. (2011). Executive (i.e., ADHD and typical controls) requires even function mechanisms of theory of mind. more advanced and sensitive quantitative EEG Journal of Autism and Developmental Disor- processing methods to reveal more subtle ders, 41, 667–678. differences between ASD and ADHD. Con- American Psychiatric Association (2000). sidering the fact that in our prior studies we Diagnostic and statistical manual of mental INDUCED EEG GAMMA ALIGNMENT IN FACIAL TASK 89

disorders (4th ed., text rev.). Washington, (2011). Differential diagnosis of autism DC: Task Force. spectrum disorder and attention deﬁcit Ashwin, C., Wheelwright, S., & Baron-Cohen, hyperactivity disorder by means of inhibitory S. (2006). Attention bias to faces in Asperger control and ‘‘theory of mind.’’ Journal of syndrome: A pictorial emotion Stroop study. Autism and Developmental Disorders, 41, Psychological Medicine, 36, 835–843. 1718–1726. Bachevalier, J., & Loveland, K. (2006). The Brock, J., Brown, C. C., Boucher, J., & Rippon, orbifrontal-amygdala circuit and self- G. (2002). The temporal binding deﬁcit regulation of social-emotional behavior in hypothesis of autism. Developmental Psy- autism. Neuroscience and Biobehavioral chopathology, 14, 209–224. Reviews, 30, 97–117. Castelli, F., Frith, C. D., Happe, F., & Frith, U. Baron-Cohen, S. (2000). Theory of mind and (2002). Autism, Asperger syndrome and autism: A review. International Review of brain mechanisms for the attribution of Research in Mental Retardation, 23, 169– mental states to animated images. Brain, 184. 125, 914–928. Baron-Cohen, S., & Belmonte, M. K. (2005). Dawson, G., Carver, L., Meltzoff, A., Autism: A window onto the development Panagiotides, H., McPartland, J., & Webb, of the social and the analytic brain. Annual S. J. (2002). Neural correlates of face and Review of Neuroscience, 28, 109–126. object recognition in young children with Baron-Cohen, S., Wheelwright, S., & Jolliffe, T. autism spectrum disorders, developmental (1997). Is there a ‘‘Language of the Eyes’’? delay, and typical development. Child Evidence from normal adults and adults Development, 73, 700–717. with autism or Asperger syndrome. Visual Dawson, G., Webb, S. J., Carver, L., Cognition, 4, 311–331. Panagiotides, H., & McPartland, J. (2004). Baruth, J., Casanova, M., El-Baz, A., Horrell, T., Young children with autism show atypical Mathai, G., Sears, L., & Sokhadze, E. (2010). brain responses to fearful versus neutral Low-frequency repetitive transcranial mag- facial expressions of emotion. Developmen- netic stimulation modulates evoked-gamma tal Science, 7, 340–359. frequency oscillations in autism spectrum de Haan, M., & Nelson, C. A. (1999). disorders. Journal of Neurotherapy, 14, Electrocortical correlates of face and object 179–194. recognition by 6 month old infants. Develop- Bas¸ar, E., Schu¨rmann, M., Bas¸ar-Eroglu, C., & mental Psychology, 35, 1113–1121. Demiralp, T. (2001). Selectively distributed Deruelle, C., Rondan, C., Gepner, B., & Tardif, gamma band system of the brain. Inter- C. (2004). Spatial frequency and face pro-

national Journal of Psychophysiology, 39, cessing in children with autism and Asperger 129–135. syndrome. Journal of Autism and Develop- Bas¸ar-Eroglu, C., Stru¨ber, D., Schu¨rmann, M., mental Disorders, 34, 199–210. Stadler, M., & Bas¸ar, E. (1996). Gamma- Di Martino, A., & Castellanos, F. X. (2003). band responses in the brain: A short review Functional neuroimaging of social cognition of psychophysiological correlates and func- in pervasive developmental disorders: A tional signiﬁcance. International Journal of brief review. Annals of New York Academy Psychophysiology, 24, 101–112. of Sciences, 1008, 256–260. Belmonte, M. K., Allen, G., Beckel-Mitchener, First, M. B., Spitzer, R. L., Gibbon, M., & A., Boulanger, L., Carper, R., & Webb, S. J. Williams, J. B. W. (2001a). Structured clinical (2004). Autism and abnormal development interview for DSM-IV-TR axis I disorders— of brain connectivity. Journal of Neuro- Patient edition (SCID—I=P). New York: science, 24, 9228–9231. New York State Psychiatric Institute. Buhler, E., Bachmann, C., Goyert, H., Heinzel- First, M. B., Spitzer, R. L., Gibbon, M., & Gutenbrunner, M., & Kamp-Becker, I. Williams, J. B. W. (2001b). Structured clinical 90 E. GROSS ET AL.

interview for DSM–IV–TR disorders— Autism and Developmental Disorders, 41, Non-patient edition (SCID—NP). New York: 512–517. New York State Psychiatric Institute. Llinas, R. R., & Steriade, M. (2006). Bursting of Frith, C. D., & Frith, U. (1999). Interacting thalamic neurons and states of vigilance. minds—A biological basis. Science, 286, Journal of Neurophysiology, 95, 3297–3308. 1692–1695. Mu¨ller, M. M., Gruber, T., & Keil, A. (2000). Frith, U. (2001). Mind blindness and the brain Modulation of induced gamma band activity in autism. Neuron, 32, 969–979. in the human EEG by attention and visual Gray, C. M., & Singer, W. (1989) information processing. International Journal Stimulus-speciﬁc neuronal oscillations in of Psychophysiology, 38, 283–299. orientation columns of cat visual cortex. Pro- Rippon, G., Brock, J., Brown, C., & Boucher, J. ceedings of the National Academy of (2007). Disordered connectivity in the autis- Sciences of USA, 86, 1698–1702. tic brain: Challenges for the ‘‘new psycho- Greimel, E., Schulte-Ruther, M., Kircher, T., physiology.’’ International Journal of Kamp-Becker, I., Remschmidt, H., Fink, G., Psychophysiology, 63, 164–172 ... Konrad, K. (2010). Neural mechanisms Rodriguez, E., George, N., Lachaux, J. P., of empathy in adolescents with autism spec- Martinerie, J., Renault, B., & Varela, F. J. trum disorder and their fathers. NeuroImage, (1999). Perception’s shadow: Long distance 49, 1055–1065. synchronization of human brain activity. Herrmann, C. S., & Demiralp, T. (2005). Nature, 397, 430–433. Human EEG gamma oscillations in neurop- Rommelse, N. N. J., Geurts, H. M., Franke, B., sychiatric disorders. Clinical Neurophysiol- Buitelaar, J. K., & Hartman, C. A. (2011). A ogy, 116, 2719–2733. review on cognitive and brain endopheno- Horrell, T., El-Baz, A., Baruth, J., Tasman, A., types that may be common in autism spec- Sokhadze, G., Stewart, C., & Sokhadze, E. trum disorder and attention-deﬁcit= (2010). Neurofeedback effects on evoked hyperactivity disorder and facilitate the and induced EEG gamma band reactivity search for pleiotropic genes. Neuroscience to drug-related cues in cocaine addiction. and Behavioral Reviews, 35, 1363–1396. Journal of Neurotherapy, 14, 195–216. Sabbagh, M. A. (2004). Understanding orbito- Keil, A., Muller, M. M., Gruber, T., Wienbruch, frontal contributions to theory-of-mind C., Stolarova, M., & Elbert, T. (2001). Effects reasoning: Implications for autism. Brain of emotional arousal in the cerebral hemi- and Cognition, 55, 209–219. spheres: A study of oscillatory brain activity Sabbagh, M. A., Moulson, M. C., & Harkness, and event-related potentials. Clinical Neuro- K. L. (2004). Neural correlates of mental

physiology, 112, 2057–2068. state decoding in human adults: An LeCouter, A., Lord, C., & Rutter, M. (2003). event-related potential study. Journal of The Autism Diagnostic Interview—Revised Cognitive Neuroscience, 16, 415–426. (ADI-R). Los Angeles, CA: Western Psycho- Sahin, I., & Yilmazer, N. (2010, May). A dis- logical Services. crete Fourier transform method for align- Lenz, D., Krauel, K., Schadow, J., Baving, L., ment of visual evoked potentials. In Duzel, E., & Herrmann, C. S. (2008). Computational intelligence in bioinformatics Enhanced gamma-band activity in ADHD and computational biology, IEEE Sym- patients lacks correlation with memory per- posium, Montreal, Canada. formance found in healthy children. Brain Schneider, W., Eschman, A., & Zuccolotto, A. Research, 1235, 117–132. (2002). E-prime user’s guide. Pittsburgh, Lerner, M., Hutchins, T., & Prelock, P. (2011). PA: Psychology Software Tools Inc. Brief report: Preliminary evaluation of the Schultz, R. T. (2005). Developmental deﬁcits theory of mind inventory and its relation- in social perception in autism: The role ship to measures of social skills. Journal of of the amygdala and fusiform face area. INDUCED EEG GAMMA ALIGNMENT IN FACIAL TASK 91

International Journal of Developmental Tallon-Baudry, C. (2003). Oscillatory syn- Neuroscience, 23, 125–141. chrony and human visual cognition. Journal Singer, W. (1999) Neuronal synchrony: A ver- of Physiology-Paris, 97, 355–363. satile code for the deﬁnition of relations. Tallon-Baudry, C., & Bertrand, O. (1999). Neuron, 24, 49–65. Oscillatory gamma activity in humans and Sokhadze, E., El-Baz, A., Baruth, J., Mathai, G., its role in object representation. Trends in Sears, L., & Casanova, M. (2009). Effect of a Cognitive Sciences, 3, 151–162. low-frequency repetitive transcranial magnetic Wechsler, D. (1999). Wechsler Abbreviated stimulation (rTMS) on induced gamma fre- Scale of Intelligence. San Antonio, TX: The quency oscillations and event-related poten- Psychological Corporation. tials during processing of illusory ﬁgures in Wechsler, D. (2003). Wechsler Intelligence autism spectrum disorders. Journal of Autism Scale for Children–Fourth Edition Integrated. and Developmental Disorders, 39, 619–634. San Antonio, TX: Harcourt.